Research On User Fine-grained Reliability And Truth Estimate Model Mobile Crowdsensing

In mobile crowdsensing(MCS)networks,users,as service nodes,play an important role in collecting crowdsensing data.However,the quality of the crowdsensing data provided by users varies to some extent depending on their perceptual capabilities and reliability.Wherein,low-quality users may provide noisy or even inaccurate data,affecting the effectiveness of the mobile crowdsensing platforms.Therefore,how to filter high-quality users through user modeling,and then accurately predict the truth value of the crowdsensing task is crucial for crowdsensing platform.This thesis proposes a general crowdsensing user fine-grained reliability and groundtruth estimate model to measure the quality of crowdsensing data.On the one hand,considering the diversity of tasks and the heterogeneity of users,a comprehensive user fine-grained reliability model is designed to measure the quality of users.Specifically,the real-time reliability of the user is evaluated according to the instantaneous factors that affect the execution of the task.,including user willingness,regional familiarity and task preference.Secondly,in the aspect of depicting historical reputation,information entropy is introduced to measure the reputation distribution of users,including the overall reputation distribution of users and the reputation distribution under different tasks.The user reliability value is initialized through the real-time reliability and historical reputation.On the other hand,using the user reliability initialization value,two kinds of truth prediction algorithms are designed based on the optimization method: the basic truth estimation algorithm and the improved truth estimation algorithm.Using the block coordinate descent method,the estimated truth value of the crowdsensing task is calculated through two steps of iterative truth value estimation and weight estimation until convergence.This thesis compares the proposed method with several other benchmark methods through extensive experiments,and uses the root mean square error and other indicators to verify and in-depth analysis of the proposed model and algorithm performance.The experimental results show that the model proposed in this thesis can effectively improve the accuracy of the prediction of the truth value of the crowdsensing task,while reducing the number of iterations of the algorithm.And in the case of malicious users in the crowdsensing system,the truth value of the crowdsensing task can still be estimated more accurately.